For Procurement

An evaluation you can run
without taking anyone's word.

Procurement's job is to make the vendor prove it. Good — this page is built for that: the standard artifacts on request, published pricing with a hard ceiling, a 30-day pilot with metrics and exit criteria defined up front, and an exit story that doesn't require a lawyer. Start with hello@turtleaicoworker.com.

What you'll want to know

The five questions
on your checklist.

Answered the way you'd want them in the vendor file: what's available, what it costs, how the evaluation runs, and how you leave.

01

Are the standard evaluation artifacts available?

Yes. A security packet — architecture overview, data-flow diagrams, current compliance status — is available on request, and we complete security questionnaires as part of the job, not a favor. The DPA is available on request with SCCs where applicable, and the whole security posture is published for inspection on the Trust Center.

The Trust Center
02

What's the pricing model?

Per-plan pricing, published in full — you can read the number before you talk to anyone. And there are no surprise usage cliffs, structurally: budget caps are hard stops enforced at runtime, with breach-date forecasting before you hit them. The invoice can't exceed a ceiling you didn't set.

See the plans
03

How do we run a structured evaluation?

With the 30-day pilot framework below: install a solution pack in about 15 minutes, define success metrics before the first run, keep approvals on everything for the first two weeks, then loosen gates only on workflows that earned it — and decide against numbers, not impressions. Exit criteria are agreed in week one, so nobody has to negotiate their way out later.

Pick a pilot pack
04

What's the exit story?

Delete means delete: remove your workspace and its data is deleted — tables, knowledge, run history and stored credentials with it. Before that, your data is exportable: audit trails, table data, the evidence you accumulated during the pilot. Leaving is a supported operation, not a negotiation.

Data handling commitments
05

Who else has evaluated this?

We won't hand you a logo wall or an arranged reference call, and we'd be suspicious of a vendor who did. What we offer instead is an evaluation that stands on inspectable evidence: a published security posture, a runtime you can gate with your own policies, an audit trail of everything the pilot did, and metrics you defined. Verify, don't take references.

The readiness guide
The 30-day pilot framework

Four weeks, metrics first, decision at the end.

The structure that keeps a pilot from becoming a permanent maybe: success metrics and exit criteria are written down in week one, and week four is a decision — either way.

WeekFocusWhat happens
Week 1Install and instrumentInstall a solution pack (~15 minutes), connect your tools, set budget caps and autonomy scopes, and write down the success metrics and exit criteria — before the first run, so the finish line can't move.
Week 2Run with approvals on everythingReal workflows, real data, every write gated by a human approval. You're measuring quality and fit while the runtime authorization floor guarantees nothing irreversible happens without sign-off.
Week 3Loosen gates on proven workflowsWhere week 2's approvals were consistently rubber-stamps, widen the autonomy scope — workflow by workflow, never wholesale. Measure what changes: throughput, error rate, escalations.
Week 4Scorecard against the metrics, then decideReview the analytics scorecard against the week-1 metrics. Pass the exit criteria: proceed with a rollout plan. Miss them: export your data, delete the workspace, and you've spent 30 days, not a budget cycle.
Hours reclaimed

Time the pilot team got back, measured against the pre-pilot baseline for the same workflows.

Autonomous vs. approved

Tasks completed autonomously versus completed under approval — the trust curve, in numbers.

Error / escalation rate

How often output needed correction or a human had to step in, from the audit trail — not from memory.

Cost per task

Per-message token and USD tracking, summed per workflow. Compare it to what the task costs today.

Exit criteria, agreed in week 1: the metric thresholds that mean "proceed", and the ones that mean "export, delete, done." If the pilot misses them, the exit is week 4 — not a renewal conversation. Governance mechanics for the approval gates: Governance · Approvals · Analytics.

What to take with you

The vendor file, mostly pre-filled.

Stakeholder one-pagers for your colleagues: IT & Security · Finance · Legal.

For your evaluation

Start the 30-day pilot.

Tell us the workflow you want to evaluate and we'll propose the pack, the metrics, and the exit criteria — in writing, before anything is installed.

Propose a pilot